Fault Diagnosis Techniques for Autonomous Underwater Vehicles

Autonomous underwater vehicles (AUVs) have become indispensable tools for oceanographic research, subsea infrastructure inspection and naval operations. Reliable fault diagnosis is critical to ensure their safety, endurance and mission success in challenging undersea environments. Faults in actuators, notably thrusters and rudders, sensors and control systems can degrade performance or lead to mission failure. Established model-based techniques generate residuals from dynamic or observer models and assess deviations via thresholding. Signal-processing methods extract features from vibration, current or acoustic signals using time-domain statistics, frequency-domain spectra or time–frequency transforms such as continuous wavelet transform. These features are then classified via statistical or machine-learning algorithms. Recent advances in data-driven approaches leverage supervised and unsupervised learning, including principal component analysis, entropy measures and support vector machines, to detect and isolate anomalies. More recently, deep-learning architectures—particularly convolutional neural networks—have been applied to spectrograms and acoustic representations for end-to-end fault identification. Hybrid strategies that fuse model-based observers with evidence-theoretic frameworks have further enhanced robustness to noise and environmental variability. Key challenges remain in obtaining representative fault data, dealing with weak or incipient faults and ensuring real-time adaptability to evolving conditions. Addressing these challenges through few-shot learning, attention mechanisms and adaptive modelling has global significance for extended ocean missions and commercial subsea operations.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Fault Diagnosis Techniques for Autonomous Underwater Vehicles publication trend

The graph below shows the total number of articles in fault diagnosis techniques for autonomous underwater vehicles across all publications each year (not limited to Nature Index journals).

Technical terms

Autonomous Underwater Vehicle (AUV): A robotic vessel capable of conducting underwater tasks without direct human control.

Thruster: A propulsive actuator used to manoeuvre or stabilise an AUV by generating thrust in water.

Convolutional Neural Network (CNN): A deep-learning model that applies convolutional filters to extract hierarchical features from data, often used for image or time–frequency analysis.

Variational Mode Decomposition (VMD): A signal-processing technique that decomposes signals into band-limited intrinsic mode functions for feature extraction.

Dempster–Shafer (D-S) evidence theory: A mathematical framework for combining evidence from multiple sources to calculate degrees of belief in hypotheses.

Meta-learning: A learning paradigm in which models are trained to adapt quickly to new tasks with minimal data by optimising across a distribution of related tasks.

References

1. A Fault Diagnosis Method for the Autonomous Underwater Vehicle via Meta-Self-Attention Multi-Scale CNN. Journal of Marine Science and Engineering (2023).
2. Enhanced Convolutional Neural Network for In Situ AUV Thruster Health Monitoring Using Acoustic Signals. Sensors (2022).
3. Fault feature extraction and fusion method for AUV with weak thruster fault based on variational mode decomposition and D-S evidence theory. Mathematical Biosciences and Engineering (2022).